I made some suggestions to Robert Wall about ways that he could improve his already quite nice Bonsai Bugs tool. As I mentioned earlier, I miss Asa’s informed and concise buildbar comments about what bugs are most interesting in each build.
Seeing Bonsai Bugs got me wondering if there was some way to determine the most interesting bugs without needing a human investigator. If Google’s shown us anything, it’s that you can use the power of the network to show you the most relevant information. What if Bonsai Bugs could somehow mark the bugs that have the most community interest and downplay those that are less important?
I came to the conclusion that there are several ways to evaluate community interest using data already in bugzilla. For example, the following criteria could be added together to devise a reasonable Community Interest Factor so that you could rank the bugs:
- Bugs with many duplicates
- Bugs with many votes
- Bugs with a large list of CCs
- Bugs with a large number of comments
- Bugs with many dependencies (or even with a parent with many dependencies)
- Bugs with certain keywords (for example, bugs fixed that have out of date milestones (Mozilla0.9), crash, dogfood, catfood, or 4xp keywords).
- Joke: Bugs that I vote for or that I am CC’ed on.
Obviously these are somewhat vague and need to be tweaked to pick up those bugs that are most relevant. You could compare with the most frequently reported bugs list or recent duplicates list as well. You could search and see if a bug is a dependency of a “Make X release not suck” bug. I imagine you could even search Google or the newsgroups for each bug number and up the Community Interest Factor if you find additional links.
I’m happy to say that Robert found the idea a good one and we’ve had a stimulating discussion about ways to refine and improve this. You could be incredibly granular and show a great number of levels of community interest, but the result is that nothing really stands out. Five seemed like too many. I suggested to him that with the right threshold only two levels (normal and interesting) might be needed, but three felt about right (normal, interesting, and imporant).
I’m looking forward to the first release of the improved Bonsai Bugs. I imagine that the algorithms will continue to be refined and this will make it more enjoyable for the community to watch the tree grow.